GLM 5.3
GLM 5.3 is served by 41 companies at prices ranging from $0.06 to $2.80 per million input tokens, a 50.9× difference for identical weights.
- cheapest host · Morph
- $0.06
- price spread across hosts
- 51×
- companies serving it
- 41
- intelligence index
- 44.8
Is it worth it?
No cheaper model matches it on quality.
It beats Kimi K3, GLM 5.2 at the same quality band, for more money.
Every host, compared
The same weights, priced and configured differently by 41 companies. Precision matters: 4-bit builds measurably degrade output and cost the same.
| Host | $/1M in | $/1M out | Cached | Context | Precision | Uptime |
|---|---|---|---|---|---|---|
| Morphcheapest | $0.055 | $5.00 | $0.060 | 1.05M | fp8 | 100% |
| Wafer | $0.079 | $4.80 | $0.059 | 1.05M | unknown | 100% |
| InferenceNet | $0.080 | $4.40 | $0.055 | 1.05M | fp4 | 100% |
| Reka | $0.140 | $4.20 | $0.105 | 262K | unknown | 99.6% |
| Makora | $0.140 | $4.40 | $0.100 | 1.05M | fp4 | 99.7% |
| Wafer | $0.170 | $4.80 | $0.160 | 1.05M | unknown | 100% |
| AkashML | $0.190 | $4.40 | $0.190 | 1.05M | fp8 | 100% |
| Sail Research | $0.200 | $3.40 | $0.150 | 1.05M | fp8 | 100% |
| Sail Research | $0.200 | $3.40 | $0.150 | 1.05M | fp8 | 100% |
| Baidu | $0.350 | $1.10 | $0.065 | 1.05M | fp8 | 100% |
| DeepInfra | $0.563 | $2.50 | $0.125 | 1.05M | fp4 | 99.1% |
| Io Net | $0.590 | $2.90 | $0.190 | 262K | fp8 | 100% |
| Inceptron | $0.600 | $1.10 | $0.200 | 1.05M | fp4 | 99.9% |
| Tenstorrent | $0.600 | $2.50 | $0.100 | 1.04M | fp8 | 96.7% |
| Novita | $0.700 | $2.20 | $0.130 | 1.05M | fp8 | 100% |
| Phala | $0.840 | $2.64 | $0.156 | 1.05M | unknown | 99.9% |
| Decart | $0.842 | $2.65 | $0.196 | 1.05M | fp4 | 100% |
| DigitalOcean | $0.910 | $2.86 | $0.169 | 1.05M | unknown | 100% |
| GMICloud | $0.980 | $3.08 | $0.182 | 1.05M | fp8 | 100% |
| SiliconFlow | $1.12 | $3.52 | $0.208 | 1.05M | fp8 | 100% |
| Alibaba | $1.19 | $3.74 | $0.238 | 1.00M | unknown | 100% |
| Friendli | $1.26 | $3.96 | $0.234 | 1.05M | unknown | 99.9% |
| Mistral | $1.40 | $4.40 | $0.140 | 1.05M | nvfp4 | 100% |
| BaseTen | $1.40 | $4.40 | $0.140 | 1.05M | fp4 | 98.6% |
| Mistral | $1.40 | $4.40 | $0.140 | 1.05M | nvfp4 | 100% |
| Crusoe | $1.40 | $4.40 | $0.260 | 1.05M | fp4 | 100% |
| PrimeIntellect | $1.40 | $4.40 | $0.260 | 1.05M | unknown | 100% |
| Venice | $1.40 | $4.40 | $0.260 | 1.00M | unknown | 100% |
| Together | $1.40 | $4.40 | $0.260 | 1.05M | unknown | 99.2% |
| Parasail | $1.40 | $4.40 | $0.260 | 1.05M | fp8 | 100% |
| Modal | $1.40 | $4.40 | $0.260 | 1.05M | unknown | 100% |
| BaseTen | $1.40 | $4.40 | $0.140 | 1.05M | fp4 | 98.6% |
| Fireworks | $1.40 | $4.40 | $0.260 | 1.05M | unknown | 99.3% |
| Cloudflare | $1.40 | $4.40 | $0.260 | 1.05M | unknown | 100% |
| AtlasCloud | $1.40 | $4.40 | $0.260 | 1.05M | fp8 | 100% |
| Z.AI | $1.40 | $4.40 | $0.260 | 1.05M | fp8 | 100% |
| Mistral | $1.54 | $4.84 | $0.154 | 1.05M | nvfp4 | 100% |
| Fireworks | $2.10 | $6.60 | $0.390 | 1.05M | unknown | 100% |
| BaseTen | $2.10 | $6.60 | $0.210 | 1.05M | fp8 | 100% |
| BaseTen | $2.10 | $6.60 | $0.210 | 1.05M | fp8 | 100% |
| Alibaba | $2.80 | $8.80 | $0.560 | 1.00M | unknown | 100% |
8 of 41 hosts serve this model at 4-bit precision (InferenceNet, Makora, DeepInfra, Inceptron, Decart, BaseTen, Crusoe, BaseTen), at the same price as full-precision hosts. See why this matters.
What it costs to run
- Marketing emails
- $2.02
- per 1,000 · short brief in, one email out
- Support replies
- $1.93
- per 1,000 · ticket plus help-centre context
- Lead enrichment
- $3.04
- per 1,000 · company page in, fields out
- Long-form drafts
- $8.87
- per 1,000 · outline in, full draft out
- Questions on docs
- $7.46
- per 1,000 · one large document, many questions
- Agent / tool loop
- $6.95
- per 1,000 · long context reused each step
Priced on the median host. Change the workload on the leaderboard to see how the ranking moves.
Specification
- Context, most hosts
- 1.05M · 1,573 pages
- Context range across hosts
- 262K to 1.05M
- Max output
- 944K
- Cache discount
- 95% off cached input
- Input types
- text
- Tool use
- supported
- Reasoning
- always on, billed as output
- Arena Elo
- 1295 across 8 categories
- Agentic Elo
- 1212
- AA Intelligence
- 44.8
Other models with the same host spread
Host pricing and precision read live from OpenRouter’s endpoint API, compiled 12 October 2026. Free to quote with attribution. See all findings.